On Sun, Jun 15, 2003 at 02:46:11AM +0900, Izumi Tsutsui wrote:
> In article <D6EF96FC-9E8C-11D7-88A6-000A957650EC@wasabisystems.com>
> thorpej@wasabisystems.com wrote:
> 
> > On Saturday, June 14, 2003, at 10:21  AM, Izumi Tsutsui wrote:
> > 
> > > There is the following comments in si.c:
> > >>> 	 * Never attempt single transfers of more than 63k, because
> > >>> 	 * our count register may be only 16 bits (an OBIO adapter).
> > >>> 	 * This should never happen since already bounded by minphys().
> > >
> > > si_minphys() in sys/arch/sun3/dev/si.c limits b_count < MAX_DMA_LEN,
> > > but why xlen is larger than MAX_DMA_LEN in si_dma_alloc()?
> > 
> > The issue here is that the file system code does not use minphys() to 
> > clamp the transfers.
> 
> Hmm, then I'm afraid there are many MI/MD SCSI drivers which
> assume each transfer size is clamped by their own minphys functions..

I don't think so, minphys was never used by the filesystem code.
However there may be other ports which uses MAXPHYS != MAXBSIZE

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 24 ans d'experience feront toujours la difference
--